作者位次 #4/5ACL 2026Findings· Findings of ACL
T hink B rake: Efficient Reasoning via Log-Probability Margin Guided Decoding
Sangjun Song · Minjae Oh · Seungkyu Lee · Sungmin Jo · Yohan Jo
摘要
Large Reasoning Models (LRMs) allocate substantial inference-time compute to Chain-of-Thought (CoT) reasoning, improving performance on mathematics, scientific QA, and tool usage. However, this introduces overthinking: LRMs often reach a correct intermediate solution, continue reasoning, and over...